pytorch实现迁移学习

发表于 | 分类于

两个主要的迁移学习场景：

  • Finetuning the convnet: 我们使用预训练网络初始化网络，而不是随机初始化，就像在imagenet 1000数据集上训练的网络一样。其余训练看起来像往常一样。
  • ConvNet as fixed feature extractor: 在这里，我们将冻结除最终完全连接层之外的所有网络的权重。最后一个全连接层被替换为具有随机权重的新层，并且仅训练该层。

参考链接：

1 https://pytorch.apachecn.org/docs/1.0/transfer_learning_tutorial.html

2 https://pytorch.apachecn.org/docs/1.0/finetuning_torchvision_models_tutorial.html

导包

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
from __future__ import print_function, division

import torch
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
import numpy as np
import torchvision
from torchvision import datasets, models, transforms
import matplotlib.pyplot as plt
import time
import os
import copy

plt.ion()   # interactive mode

os.environ["CUDA_VISIBLE_DEVICES"] = "9"
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

加载数据

我们将使用 torchvision 和 torch.utils.data 包来加载数据。

我们今天要解决的问题是训练一个模型来对 蚂蚁 和 蜜蜂 进行分类。我们有大约120个训练图像，每个图像用于 蚂蚁 和 蜜蜂。每个类有75个验证图像。通常，如果从头开始训练，这是一个非常小的数据集。由于我们正在使用迁移学习，我们应该能够合理地推广。

该数据集是 imagenet 的一个非常小的子集。

注意

从 此处 下载数据并将其解压缩到当前目录。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
# Data augmentation and normalization for training
# Just normalization for validation
data_transform = {
    'train': transforms.Compose([
        transforms.RandomResizedCrop(224),
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ]),
    'val': transforms.Compose([
        transforms.Resize(256),
        transforms.CenterCrop(224),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ])
}

data_dir = 'hymenoptera_data/'
image_datasets = {x: datasets.ImageFolder(os.path.join(data_dir, x), 
                                          data_transform[x]) 
                  for x in ['train', 'val']}
dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=4,
                                             shuffle=True, num_workers=4)
              for x in ['train', 'val']}
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'val']}
class_names = image_datasets['train'].classes
print(dataset_sizes)
print(class_names)
{'train': 244, 'val': 153}
['ants', 'bees']

可视化图像

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
def imshow(inp, title=None):
    """Imshow for Tensor"""
    inp = inp.numpy().transpose((1,2,0))
    mean = np.array([0.485, 0.456, 0.406])
    std = np.array([0.229, 0.224, 0.225])
    inp = inp * std + mean
    inp = np.clip(inp, 0, 1)
    plt.imshow(inp)
    if title is not None:
        plt.title(title)
    # plt.pause(0.001)
    
# get a batch of training data
inputs, classes = next(iter(dataloaders['train']))

# make a grid from batch
out = torchvision.utils.make_grid(inputs)

imshow(out, title=[class_names[x] for x in classes])

训练模型的通用函数

函数有以下功能：

  • 训练模型
  • 调整学习率
  • 保存最佳的学习模型

函数中, scheduler 参数是 torch.optim.lr_scheduler 中的 LR scheduler 对象

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
def train_model(model, criterion, optimizer, scheduler, num_epochs=25):
    since = time.time()
    
    val_acc_history = []
    
    best_model_wts = copy.deepcopy(model.state_dict())
    best_acc = 0.0
    
    for epoch in range(num_epochs):
        print('Epoch {}/{}'.format(epoch + 1, num_epochs))
        print('-'*10)
        
        # each epoch has a train and valid phase
        for phase in ['train', 'val']:
            if phase == 'train':
                scheduler.step()
                model.train()    # Set model to training mode
            else:
                model.eval()     # Set model to evaluate mode
            
            running_loss = 0.0
            running_corrects = 0
            
            # Iterate over data.
            for inputs, labels in dataloaders[phase]:
                inputs = inputs.to(device)
                labels = labels.to(device)
                
                # zero the parameter gradients
                optimizer.zero_grad()
                
                # forward
                # track history if only in train
                with torch.set_grad_enabled(phase == 'train'):
                    outputs = model(inputs)
                    _, preds = torch.max(outputs, 1)
                    loss = criterion(outputs, labels)
                    
                    # backward + optimize only if in training phase
                    if phase == 'train':
                        loss.backward()
                        optimizer.step()
                        
                # statistics
                running_loss += loss.item() * inputs.size(0)
                running_corrects += torch.sum(preds == labels.data)
                
            epoch_loss = running_loss / dataset_sizes[phase]
            epoch_acc = running_corrects.double() / dataset_sizes[phase]
            
            print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))
            
            # deep copy the model
            if phase == 'val' and epoch_acc > best_acc:
                best_acc = epoch_acc
                best_model_wts = copy.deepcopy(model.state_dict())
            if phase == 'val':
                val_acc_history.append(epoch_acc)
            
        print()
        
    time_elapsed = time.time() - since
    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, 
                                                       time_elapsed % 60))
    print('Best val Acc: {:.4f}'.format(best_acc))
    
    # load best model weights
    model.load_state_dict(best_model_wts)
    return model, val_acc_history

可视化模型函数

用于显示少量图像预测的通用功能

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
def visualize_model(model, num_images=6):
    was_training = model.training
    model.eval()
    images_so_far = 0
    
    fig = plt.figure()
    
    with torch.no_grad():
        for i, (inputs, labels) in enumerate(dataloaders['val']):
            inputs = inputs.to(device)
            labels = labels.to(device)

            outputs = model(inputs)
            _, preds = torch.max(outputs, 1)
            
            for j in range(inputs.size()[0]):
                images_so_far += 1
                ax = plt.subplot(num_images//2, 2, images_so_far)
                plt.tight_layout()
                ax.axis('off')
                ax.set_title('true: {} ==> predicted: {}'
                             .format(class_names[labels[j]], class_names[preds[j]]))
                imshow(inputs.cpu().data[j])
               
                
                if images_so_far == num_images:
                    model.train()
                    return
    
    model.train(model=was_training) # 似乎有点问题

微调卷积网络

加载预训练模型并重置最终的全连接层

1
2
3
4
5
6
7
8
9
10
11
12
13
model_ft = models.resnet18(pretrained=True)
num_ftrs = model_ft.fc.in_features
model_ft.fc = nn.Linear(num_ftrs, 2)

model_ft = model_ft.to(device)

criterion = nn.CrossEntropyLoss()

# Observe that all parameters are being optimized
optimizer_ft = optim.SGD(model_ft.parameters(), lr=0.001, momentum=0.9)

# Decay LR by a factor of 0.1 every 7 epochs
exp_lr_scheduler = lr_scheduler.StepLR(optimizer_ft, step_size=7, gamma=0.1)

训练和评估

1
model_ft, val_acc_history = train_model(model_ft, criterion, optimizer_ft, exp_lr_scheduler)
Epoch 1/25
----------
train Loss: 0.6349 Acc: 0.7049
val Loss: 0.5844 Acc: 0.7451

...

Epoch 25/25
----------
train Loss: 0.3083 Acc: 0.8484
val Loss: 0.2406 Acc: 0.9085

Training complete in 29m 38s
Best val Acc: 0.9346

1
visualize_model(model_ft)

ConvNet 作为固定特征提取器

在这里，我们需要冻结除最后一层之外的所有网络。我们需要设置 requires_grad == False 冻结参数，以便在 backward() 中不计算梯度。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
model_conv = models.resnet18(pretrained=True)
for param in model_conv.parameters():
    param.requires_grad = False

# Parameters of newly constructed modules have requires_grad=True by default
num_ftrs = model_conv.fc.in_features
model_conv.fc = nn.Linear(num_ftrs, 2)

model_conv = model_conv.to(device)

criterion = nn.CrossEntropyLoss()

# Observe that only parameters of final layer are being optimized as
# opposed to before.
optimizer_conv = optim.SGD(model_conv.fc.parameters(), lr=0.001, momentum=0.9)

# Decay LR by a factor of 0.1 every 7 epochs
exp_lr_scheduler = lr_scheduler.StepLR(optimizer_conv, step_size=7, gamma=0.1)

训练和评估

在CPU上，与前一个场景相比，这将花费大约一半的时间。这是预期的，因为不需要为大多数网络计算梯度。但是，前向传递需要计算梯度。

1
2
model_conv, val_acc_hietory = train_model(model_conv, criterion, optimizer_conv,
                         exp_lr_scheduler, num_epochs=10)
Epoch 1/10
----------
train Loss: 0.6322 Acc: 0.6148
val Loss: 0.2261 Acc: 0.9346

...

Epoch 10/10
----------
train Loss: 0.4004 Acc: 0.8197
val Loss: 0.1821 Acc: 0.9477

Training complete in 6m 5s
Best val Acc: 0.9477

1
2
3
4
5
6
7
8
9
10
11
hist = [h.cpu().numpy() for h in val_acc_hietory]

plt.figure()

plt.title("Validation Accuracy")
plt.xlabel("Training Epochs")
plt.ylabel("Validation Accuracy")
plt.ylim((0,1.))
plt.xticks(np.arange(1, 11, 1.0))
plt.plot(np.arange(1,11),hist,label="Pretrained")
plt.legend()

1
2
3
4
visualize_model(model_conv)

plt.ioff()
plt.show()